ReadingStar

Reading and pronunciation difficulties pose significant challenges for many individuals, including children, non-native speakers, and people with accessibility needs. Existing speech recognition tools often struggle with accents and require cloud-based processing, raising privacy concerns and limiting accessibility in offline environments. There is a need for an interactive, real-time, and offline solution that enhances speech recognition and reading fluency using AI.

We have created ReadingStar, an AI-powered, offline read-along and sing-along app using Intel OpenVINO's optimisation pipeline and the OpenAI Whisper transcription model. It transcribes lyrics in real time, highlights progress as users speak, and offers feedback on pronunciation. Running on an Intel NPU ensures fast, private, and efficient speech processing with multiple difficulty modes for different learning needs.

The interactive elements of the application are designed to encourage enthusiastic engagement, igniting users' passion for singing so they may train and practice their speech, whilst increasing both enunciation and reading proficiency.